Qoris Launches Knox: Multi-Surface Security Policy Engine for AI Coding Agents
Key Takeaways
- ▸Knox provides independent, policy-driven security enforcement for AI agents across multiple deployment surfaces, complementing rather than replacing model-level safety measures
- ▸Multi-form distribution strategy allows organizations to adopt Knox at different points—from local development (CLI/plugins) to production automation (enterprise runtime)
- ▸Enterprise version adds production-grade features like approval workflows, shared memory governance, and audit pipelines for AI workers running 24/7 across business operations
Summary
Qoris has launched Knox, a security policy engine that governs AI agent tool calls before execution. Available in five deployment forms—standalone CLI, Node library, and native plugins for Claude Code, Cursor, and OpenAI Codex—Knox provides independent policy enforcement separate from model-level safety measures. The offering includes two tiers: Developer Knox (free, open source) for local developer environments, and Qoris Runtime Knox (enterprise) for production AI workflows running across sales, operations, compliance, and support teams.
Knox enables granular security controls including real-time blocking of dangerous operations, automatic audit logging, prompt injection scanning, and self-protection against configuration tampering. A key distinction: CLI and library versions provide inspection and policy decisions, while IDE plugin versions add real-time enforcement at execution time. The system intercepts 11 hook events across different development surfaces, sharing one unified rule set across all forms.
Editorial Opinion
Knox fills a critical gap in AI development infrastructure: as AI agents become capable of executing code and making system changes autonomously, model-level safety is necessary but insufficient. Independent policy enforcement—separate from the model's decision-making—enables true defense-in-depth security architecture. The multi-surface approach is particularly smart, allowing organizations to adopt security controls at the point that matches their workflow maturity, from individual developers to enterprise-scale production automation.
